A Raspberry Pi is extreme, but again, I’m not talking about a few block size doublings. “Visa scale” is on the order of 1000x Bitcoin’s current capacity.
The other argument I’m familiar with is longer block propagation times lead to more orphan blocks, which is a centralization pressure.
EDIT: responding to your edits:
> layer 2 payment solution which essentially will force transactions to route through centralized middlemen
A payment system's fees need to be 1-2% at most to match credit cards. I remember fees going as high as $80 for a transaction back in the day, so that means you could only reasonably use bitcoin for a minimum of $1000-$10,0000 when fees were the worst. But in reality, Bitcoin can be .000001% (I put a random number of zeroes don't take it literally).
Additionally you're trying to "protect the blockchain" by having people never able to use it. Surely you see the absurdity.
If Bitcoin ever hit Visa scale, there'd be no problem with only well-capitalized miners maintaining a full chain. It's really not an issue, but frankly you should cross that bridge when we get to it anyway. In actuality Bitcoin was hard limited at 3-7 transactions/sec for no reason whatsoever.
The other argument I’m familiar with is longer block propagation times lead to more orphan blocks, which is a centralization pressure.
EDIT: responding to your edits:
> layer 2 payment solution which essentially will force transactions to route through centralized middlemen
1. Layer 2 solutions aren’t necessarily centralized
2. Even if they are all centralized I don’t believe it’s worth sacrificing decentralization in the base layer to support small payments